Few-Shot Learning for Rooftop Detection in Satellite Imagery

Deep Learning Tutorial

Giorgio Coppala, Nadine Daum, Elena Dreyer, Nico Reichardt

Tentative Structure (delete me)

problem setting (policy relevance)
dataset geneva
model & methods
prototypical Networks
main notebook in detail
results to expect
wrap-up

Problem Setting

Cities need accurate rooftop maps to plan and scale solar PV installations
Manual rooftop labeling is slow and costly
Every city looks different → traditional models do not generalize well

Idea:

Few-shot learning makes segmentation possible with only a handful of labeled examples

Dataset: Geneva

Satellite Images: High-resolution RGB satellite images of Geneva
Size: 1,050 labeled image-mask pairs
Task: Binary segmentation masks (rooftop vs background)
Geographic splits: 3 grids/ neighborhoods (1301_11, 1301_13, 1301_31)
Image size: 250x250 pixels
Categories: Industrial, Residential

Model & Methods

Data Preprocessing
Model Architecture
Few-Shot in a Nutshell (modified figure from paper)
Few-Shot in implementation (ntoebook reference/ pseudocode for logic?)
Training strategy
Loss function
Evaluation metrics

Prototypical Networks

high-level schematic (support → prototype → similarity → segmentation)
literature reference: SRPNet

Main Notebook in Detail

how deep should we go?

lets discuss that regarding time

(presentation should be 10 minutes, followed by 5 minutes of Q&A)

Expected Results

Show performance for 1-shot / 5-shot / full-data comparison
Show predicted masks

Open to Discuss:

strengths
weaknesses
failure cases (shadows, tiny rooftops)

Wrap-Up: GitHub Repo

insert more from discussion + memo here

What we have so far:

insert bullet point here
insert bullet point here

What we still need to finalize:

insert bullet point here
insert bullet point here

Questions to discuss in class/ lynn

insert bullet point here